A method for Constructing a Reference Image Database
to Assist with Design Process. application to the
Wooden Architecture Domain.
Sabrina Kacher, Gilles Halin, Jean-Claude Bignon, Pascal Humbert
To cite this version:
Sabrina Kacher, Gilles Halin, Jean-Claude Bignon, Pascal Humbert. A method for Construct-ing a Reference Image Database to Assist with Design Process. application to the WoodenArchitecture Domain.. IJAC, International Journal of Architectural Computing, 2005, 3 (2),pp.227–243. <10.1260/1478077054214415>. <halshs-00262180>
HAL Id: halshs-00262180
https://halshs.archives-ouvertes.fr/halshs-00262180
Submitted on 31 Mar 2008
HAL is a multi-disciplinary open accessarchive for the deposit and dissemination of sci-entific research documents, whether they are pub-lished or not. The documents may come fromteaching and research institutions in France orabroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, estdestinee au depot et a la diffusion de documentsscientifiques de niveau recherche, publies ou non,emanant des etablissements d’enseignement et derecherche francais ou etrangers, des laboratoirespublics ou prives.
A method forConstructing a ReferenceImage Database to Assistwith Design Process.
Application to theWooden ArchitectureDomainSabrina Kacher, Gilles Halin, Jean-Claude Bignon
and Pascal Humbert
227issue 02, volume 03international journal of architectural computing
A method for Constructing a Reference Image
Database to Assist with Design Process.
Application to the Wooden Architecture DomainSabrina Kacher, Gilles Halin, Jean-Claude Bignon and Pascal Humbert
Designing architectural projects requires the
introduction of references, because design is an
activity oriented towards a result which does not yet
exist. If we summarise the current categories used in
Artificial Intelligence to characterise the different
forms of reasoning, we are able to consider that
design is more the concern of the induction or the
abduction mechanism than the deduction mechanism.
Moreover, the main characteristic of the designer’s
activity is to work towards non-routine situations
with the use of many references. In this paper we will
present method principles to construct a reference
image database.These references will enable the
designer to further in solving the design problem.To
illustrate these reference usage, we choose
photographic images belonging to the wooden
construction domain We also present at the end of
the paper an experiment which aims to evaluate the
real help that this reference image database can bring
to designers during their creation task.
228
1. Introduction
Design requires research into ideas [1] and documentation to help
designers in their creation work. It is a domain where the use of pictures
(drawings, photograph, etc) is essential because the information transmitted
by images is often easier to understand than the one transmitted by texts.
The fact is that an image requires less interpretation than a text.The
information transmitted by image (element shape, colour, light, etc) is
already “put in shape” and can be more easily integrated into the design
process. For this we have chosen a specific kind of image to illustrate the
references for the designer which is “photographic” images.
Photographic image is widely used as an information support in the
design field because it can convey a large amount of information that will
help the designer to solve his design problem in a minimum of time. It also
illustrates several elements in different ways and the designer tends to put
at the same level of cognition the object and the image of the object.
The particularity of the design process is that it brings together two
distinct operations where image use is essential.The first one concerns
problem formulation (image allows the user to express his design question
and enables him to advance in his problem formulation).The second one
concerns the problem’s solution (image illustrates several potential solutions
to the design problem) [2].These operations can be done thanks to mental
images that the designer creates in his mind.
This paper tackles the particular question of the principles to follow in
order to better construct a reference image database that will help the
designer during the design process.We will also present a concrete
application of the proposed method to a particular domain which is the
wood construction domain. Finally we will present a concrete example of
how our method once implemented in a suitable search engine can help
designers during their design process.
2.The method principles
Our work consists of presenting a method to better describe, index and
access images.This method is oriented toward a semantic description of
images. Our proposition is not to suggest help to the designers who want
to index their own image database.These are indexation principles that will
be used in an information resource centre to offer an information service
to designers.To construct the database we suggest following these
principles:
• Defining the context of the work.
• Defining the structure of the chosen domain.
• Defining a way to quantify the visual importance of an element.
• Identifying a specific research process.
229A method for Constructing a Reference Image Database to Assist with Design Process.
Application to the Wooden Architecture Domain
2.1. Defining the context of the work
Even if we have decided to focus on the architecture domain, the person
that will define and propose a semantic indexation method has to reduce
and limit the field to be described. It’s a necessary task because architecture
is a wide-ranging domain and the objective is to define a vocabulary which
aims to represent and describe as much as possible a precise domain [3].
Then this definition of the context will direct all the work to follow.
Moreover all domains possess various sub-domains. For example in
architecture we have to define description, construction, design, social and
urban ways amongst others.To succeed with this method we have to define
as much as possible the domain where we work.
2.2. Defining the structure of the chosen domain
Nowadays there are several fields of research being undertaken which aim
at describing a domain in words.There are a lot of vocabularies applied to
describe various domains (ontologies, thesauruses, lexicon, etc).The most
widespread kind of existing vocabularies are “thesauruses” [4].These
thesauruses possess deep structure or hierarchy.The extents of these
hierarchies meant that it took a very long time to find the suitable word to
describe an element and then induce a wrong description of the document.
In our case we used a proposition made by Eleanor Rosch [5] to classify the
natural categories in the cognitive psychology domain. She suggested
classifying all the objects of the real world on only three hierarchical levels.
She explained that the categorisation could be done well starting from the
intermediate level.This level includes the name of the prototype of all the
current objects of the domain to be categorised.This level can include in
the field of furniture,“chair”,“table”, etc., or in the fruit domain,“apple”,
“orange”, etc.To construct the two other levels we have to respect some
rules.The process of generalisation can be carried out by identifying the
attributes of the objects named in the previous level. For example to create
the superior level of “apple” and “orange” we can create the “fruit
category”.This category includes the shared properties of an “orange” and
an “apple”.To create the inferior level Rosch proposes to add a specificity to
the term of the current level. For example, an “office chair” could be
different from a “kitchen chair”.
2.3. Defining a way to quantify the visual importance of anelement
To describe an image in a better way we have to organise the terms inside
the image indexations.Then we focus only on the description of what is
visually expressible by our thesaurus words in images. In our case the visual
parts concern the physical parts of architectural realisations.We propose a
classification or an organisation of the terms used for image indexation.This
230 Sabrina Kacher, Gilles Halin, Jean-Claude Bignon and Pascal Humbert
classification depends on the visual significance of the illustration and is
related to the angle from which the picture is shot. In some images an
architectural built work should be easier to identify than in others.This
happens specifically for junction elements which require a close-up shot to
be recognise.
2.4. Identifying a specific research process
To make the problem formulation and solution easier for the designer we
have defined a search mode adapted to the designer’s state. Nowadays, the
common access to images proposed by existing system, is the search mode
using text.This mode requires using text to express a need. But in the
architectural design process the needs are rarely precise and expressible in
words. For this we proposed a particular kind of research mode which is
the search with images.This kind of research uses images to access other
images.We feel that this search mode is more adapted to the designer need
during design activity. In our case the designer will use only images to
formulate his need. Our system mode doesn’t require text to formulate a
need.We consider this search mode more adapted to the state of need
which is imprecise and fuzzy especially in the earlier phases of the design
process (figure 5).
3.Application to a specific domain
To validate the relevance of the proposed method we applied the
enumerated principles to a concrete example. For this we chose a particular
sub-domain of architecture, which is architecture built in wood.Then,
photographic images that we work with illustrate buildings or parts of these
buildings belonging to the wood construction domain.
According to the principles enumerated above, we will propose an
indexation process in four steps as follows ;
3.1. Step (1): defining the context of the work
At the beginning, we defined a vocabulary for image indexing. Defining a
vocabulary results from the analysis of images to index.
There are two main families of image describers:
• The “Offness” family which is related to the directly extractive
information from the image. Its divided in two parts:
- The graphic content outline, texture, shape, etc. [7]
- The semantic content relative to the elements represented directly
on the image.
• The “Aboutness” family which is related to describe the
complementary information but not directly extractive information
231A method for Constructing a Reference Image Database to Assist with Design Process.
Application to the Wooden Architecture Domain
from images. Its divided in two parts:
- The analysis content relative to the name of the project, the type of
the project, the geographical situation, etc.
- The symbolic content that is located on a more elevated
abstraction level such as opacity, permeability, monotony, etc.
For this application example, we decided to focus on the part
concerning the information related to the elements represented in the
picture and the one related to the project.
In order to meet efficiently the user’s expectations and to reduce the
ambiguity resulting from the multi-interpretation of the image (polysemy),
we defined four families (Figure 1) of image describers [8]:
• The architectural realisation family; includes the name of the category
to which the architectural element illustrated belongs (school, single-
family dwelling, …).
• The built-works family; indicates every physical part of an
architectural realisation which has a particular function (post, beam,
window, ....).
• The material family; includes every wood material and its by-products
(species, glued-laminated, …).
• The products family; includes any component aimed at protecting and
decorating wooden elements (fungicide, impregnation, …).
3.2. Step (2): defining the structure of the chosen domain
The vocabulary created has to enable the formalisation of the more or less
homogeneous language-practices of the professionals in the domain [9].This
� Figure 1. Image families of
describers
232 Sabrina Kacher, Gilles Halin, Jean-Claude Bignon and Pascal Humbert
point is important because these professionals have the same scientific and
technical references.To better represent the described domain, we
structured each family of image describers, following the E.Rosch [5]
categorisation which structures in 3 hierarchical levels real world objects.
The categorisation is carried out from the intermediate level called “current
level”.The definition of each level is made following rules (Figure 2) of
generalisation and specialisation [10].The three levels are as follows [11]:
The current level (CL)
It represents the level from where the language is structured.This level is
also called a basic level because it contains the most important concepts in
the language of the domain we want to describe and therefore, the one we
learn in the first place.This level corresponds to the designer’s problem
when the problem is of a general nature. Its contains the name of
“prototype” or of the most representative member of the elements that
have properties in common. It includes the name of concrete elements
illustrated in the images e.g. post, beam, window, door, etc.
The superior level (SL)
It corresponds to the fuzzy demand of the user and guides him during a
search towards a more precise and more specific level.This level represents
a more general level than the current level.Terms belonging to this level
allow identifying the shared attributes of the architectural concepts
belonging to the current level. For example, a “post” is a “vertical structure
element”, and a “wall” is also a “vertical structure element”.The shared
attributes for both these elements are “structure” and “vertical”. So the
superior category will include these attributes and will be called “ vertical
structure system” category.
The inferior level (IL)
The terms belonging to this level allow a more concrete identification of
the architectural concepts illustrated in the images than those of the
previous levels.The concretisation corresponds to the user’s demand during
his search for precise architectural solutions to his design problems.Terms
belonging to this level are used for indexing images and contain only one
degree of extra specificity, compared to those of the current level.This
specificity is obtained by the research of characteristics such as:
• The shape;
Corresponds to the “outline” of an object.This characteristic has
two levels. i) the shape of the architectural element itself e.g. Post →
Circular post and ii) the unit used e.g. indoor flooring → panels.
233A method for Constructing a Reference Image Database to Assist with Design Process.
Application to the Wooden Architecture Domain
• The constructive mechanism;
Corresponds to the way the architectural element is arranged,
according to predetermined rules of construction. E.g.Wall → Half-
timbered wall
• The functional mechanism;
Corresponds to the way that the architectural element is used to
fulfil its characteristic role. E.g.: Cladding → Shingle cladding
• The number;
Corresponds to the multiplicity or to the uniqueness of a single
architectural element. E.g. Structure → Successive porticos
• The orientation;
Corresponds to the direction of the element with regard to the
horizontal / vertical or intermediate axes. E.g. Sun-breaker →
Horizontal sun-breaker.
• The full-empty device;
Corresponds to the extent to which the architectural element has
been opened up with the aim of letting in daylight or not. E.g.
Partition → Honeycomb wall.
3.3. Step (3): defining a way to quantify the visual importance ofan element
Describing a built-work element illustrated by an image with a vocabulary
term is not enough.An element can be visually more important in one
photographic image than in another one.This variability of importance is
related to the shape of the element and also depends on the angle from
which the view is shot (Figure 3).We organised terms used to index each
image according to its visual importance.This importance must be quantified
� Figure 2.The vocabulary
construction.
234 Sabrina Kacher, Gilles Halin, Jean-Claude Bignon and Pascal Humbert
Post WallIntersection = generalisation
Extension = specification
SLCL CL
CL CLSL
A1Punctual
A2Length
A3Structure
A4Vertical
B1Area
B2Surface
B3Structure
B4Vertical
to be controlled.We therefore identified visual properties that will enable
us to evaluate and quantify this significance [12].
We defined five image properties (Figure 4) [6]:
The occupied area
Depends on whether the illustration of the real architectural element
occupies a large area in the image or not.Thus, the representation of the
element is in a prominent position if it occupies a larger visual area than the
other elements which surround it.
The archetype likeness
Depends on the likeness of the illustration of the real architectural element
to the ideal model shared by professionals belonging to the same domain. If
the representation of the element keeps the structural and spatial
properties of the real objects so allowing the viewer to identify the element
appearing in the image, the element will be easier to recognise [13].
The contrast with the background
Depends on the capacity of the illustration of the real element to emerge
from the rest of the image.The element will be in a conspicuous position if
its representation contrasts strongly with the rest of the image (colour,
light, ...).
The focus
Depends on the position of the illustration of the real architectural element
in the image. If it occupies the centre of the image (diagonal junction), the
element should be more obvious than the rest of the elements illustrated.
The completeness
Every illustration of a real element shows only a part of it.This graphic-
property depends on whether the illustration of the object represents the
semantic features or not [14].This allows a viewer to identify the
represented element.
� Figure 3.The organisation of the
indexation terms
235A method for Constructing a Reference Image Database to Assist with Design Process.
Application to the Wooden Architecture Domain
All these graphic properties do not possess the same impact.Their
impact depends on the kind of architectural element which is illustrated in
images.Thus, we group all the architectural elements named in the
thesaurus according to the Ching [15] classification (planar elements, linear
elements and punctual elements).This classification is oriented toward a
classification of architectural elements according to their geometric
features.To evaluate the real importance of each graphic property related to
the geometric type of the element illustrated we carried out an experiment.
This experiment has allowed us to propose calculation rules that will
facilitate the terms classification task.The objective of this part is to enable
the system to propose semi automatically a value that will be associated to
a rank for every term used for the indexation [11, 12].This is carried out
according to the kind of built work to which the element indexed belongs.
The experiment allows us to extract some observed values from the Chi-
square [16] table:
• Planar elements: for planar elements the “occupied area” graphic
property represents more than 72% of the chi-square value. It
corresponds to the estimation of the importance that the property
area has for the planar element.
• Linear elements: for example, the image in illustrates both “post” and
“cladding”.These architectural elements were grouped in two
different categories according to their geometric properties.A post
belongs to linear elements and cladding belongs to surface elements.
To follow the differences between the importance of these
properties we decided to give a more important value to the most
important graphic property.
� Figure 4.The graphic properties
236 Sabrina Kacher, Gilles Halin, Jean-Claude Bignon and Pascal Humbert
The goal of this part is to define an algorithm which enables the system
to propose semi automatically a weighted value for every term used to
index images. Our proposition consists of an association of a double value
to the first most important graphic property for each family.The value
obtained with the application of these mathematical rules will constitute the
weighted value of the term.This weighted value will be considered as a basis
to classify the indexation terms. Consequently, this part allows us to
propose steps to facilitate the indexer’s task when he associates a rank with
every term used for the semantic indexation of images.The steps are as
follows:
1. Firstly, the indexer must identify the element represented by the
image that will be described.
2. Secondly, for each selected element, its representation will be
compared to the experiment results.This will enable the system to
propose rates to classify the indexed terms.
3. Thirdly, the indexer will decide if she/he will validate or refuse the
ranks proposed by the system during the indexation process.
3.4. Step (4): identifying a specific research process
The research process we used is that of the interactive and progressive
research system [6] by image which has been developed by the MAP-CRAI1.
The defined vocabulary will be inserted into the system developed by the
MAP-CRAI in order to better meet the user’s needs.A method of relevance
feedback is used by the system to propose new images for the user’s
queries.The indexing document is represented by a weighted vector of
thesaurus terms according to the term classification of the image
indexation.A vectorial matching model is then used between the query and
the indexing document.The results of this matching will be given as an
ordered list of images representing the user’s choices.
In the previous section, we proposed a particular way to access images
[17].This access mode is based on a visual formulation of the designer’s
need.The user will only manipulate images during his research. Images will
be presented in a mosaic layout, and the user will select, reject, or not give
an opinion on every image that he visualises.The selection is obviously
oriented by the design problem. For example if he searches for windows, all
the images which illustrate window will be selected by the user, and the
others will be rejected or without an opinion.The objective of this kind of
selection enables the user to concentrate only on his design task.Therefore,
when the designer expresses a need by text, it means that he expresses a
mental image in words. In this case, the user translates graphic information
into textual information. He uses part of his “mental energy” on another
237A method for Constructing a Reference Image Database to Assist with Design Process.
Application to the Wooden Architecture Domain
1“MAP-CRAI” Architecture and landscape Modeling – Research Center in Architecture and
Engineering,Architecture School of Nancy.
task, which is “translation”.Also, several research works have shown that
mental images are easier to produce and to modify starting from images.
For this reason we avoided using another kind of expression for the
designer (Figure 5).
4. Employing our method
Once all these steps have been completed, indexers can use our method to
construct an image database, and designers can use the research system to
find image references.
4.1.The indexer usage
The indexer has to:
• Select an image to index.
• Visualise the image and identify the elements to describe.
• Use the built thesaurus and select only from the inferior level-terms
which are the most suitable for describing the element identified in
the image.
• Associate the selected terms with the image.
• Analyse the classification proposed by the system. For each selected
element, its representation will be compared to the results obtained
by this experiment.According to the results obtained though this
experiment, the system proposes rates to weight the indexed terms.
• Validate or refuse the rates proposed by the system during the
indexation process.
• Register the indexation operation.
� Figure 5.The interface adapted to
image research
238 Sabrina Kacher, Gilles Halin, Jean-Claude Bignon and Pascal Humbert
The user selects,rejects or does
not give anopinion
4.2.The designer usage
When the designer has a design problem he will use the system proposed.
At the beginning he has to identify his design problem. He is not obliged to
define it in words.The designer can draw it or express it another way.The
efficiency of our method is based on a cognitive exchange process between
the mental images of the designer and the real photographic images
retrieved by the system.This happens in three steps.
• The first stage: this step concerns the state just before visualisation
of the images. In this part, the designer possesses more or less vague
notions and creates mental images according to his advancing state in
the problem solution process.The images which come to mind are
fuzzy, so he queries the system.
• The second stage: at the beginning of the research, the system
presents a mosaic of images, which were selected randomly or
thematically.The user selects or rejects the images which are
presented to him, according to his judgement on their relevance.The
system analyzes these choices, and presents to the user a new
sample of images which corresponds better to the user’s needs.
• The third stage:The last part of the cognitive process happens after
the first visualisation of the images proposed by the system. In this
part, the user shows this new sample of images and makes a
permutation by replacing the elements composing his own mental
images with those visualised in the real images. It operates a
deformation and an adaptation of his mental images, and generates
new images.This stage should allow the user to reformulate the
problem, or to start a new problem which has not appeared before,
or even to solve the design problem.
5.The final experimentation
To validate the relevance of our method, we carried out a final
experimentation with ten designers in a design situation.We gave every
subject a design problem and asked him to propose a timber structure
covering a swimming pool with 25C50 meter size.
5.1.The protocol
The experimentation was carried out in four stages.
Stage N°1
At this first stage we asked the designer to draw the first sketch (Figure 6).
239A method for Constructing a Reference Image Database to Assist with Design Process.
Application to the Wooden Architecture Domain
We can clearly identify that this first sketch is relatively purified.This
drawing includes a group of similar vertical and horizontal elements.The
lines representing these elements are simple.We can also notice that the
shape of the roof is a sloped one and the proposed framework consists of a
succession of porticos with a small roof overhang.
Stage N°2
During this stage, we asked the subject to start the research, and the first
sketch was taken away from the subject.
Stage N°3
During the third stage, the subject selects, rejects or does not give any
opinion on every image presented to him by the system.The last mosaic
visualised is presented in the next figure (Figure N°7).We can see that this
mosaic illustrates various built work elements such as joist, sun breaker,
wind bracing, posts, etc.
Step N°4
The last stage was performed by asking the subject to redraw his early
design problem (Figure 8).
5.2. Conclusion for subject N°1
For this subject, we can clearly identify that the visualised images enable him
to add information related to his design’s problem.The final sketch looks
more completed and loaded than the first sketch.The inclusion of these
� Figure 6. First sketch of subject
N°1.
240 Sabrina Kacher, Gilles Halin, Jean-Claude Bignon and Pascal Humbert
Thin lines
additional visualised peaces of information and the integration of this
information takes place during this design exercise.We can also clearly
notice that the subject added a caption or textual comments to his final
sketch with some details about the material used or the junction types
used.At the end of the experiment we asked the subject a series of
questions.These questions were related to the identification of images
which influenced his final solution for the design problem.This first subject
said “at the beginning I had not idea, but this image (centre of the mosaic) gave
me the sun breaker idea” (Figure 9).
� Figure 7. Last mosaic visualised by
subject N°1
� Figure 8. Final sketch of subject
N°1.
� Figure 9.The image from the mosaic
which inspired the sun breaker idea.
241A method for Constructing a Reference Image Database to Assist with Design Process.
Application to the Wooden Architecture Domain
Sun breaker
Wind bracingJoist
Post
Sun breaker (lower orupper part)
Large thickness of thefront (see Rémilly High
school) or front overhang(see Hanovre)
Portico (see Biennestyle).
Roof overhang, sunbreaker type-Bienne +
Sartrouville project
Thicken lines
Textual commentaries
And when we asked him what the research into built work elements
gave him during his design process he answered “the visualised images
illustrate real architectural realisations, so we can very quickly get an idea of
what the built work in question will look like once integrated into the
project (we have an example of the use of this built work)”.
6. Conclusion
In this article we present four method principles for defining a way to
describe index and access images.This method aims to facilitate the
construction of a photographic image database that will play the role of a
reference tool for the designer during his design activity.These principles
consist of defining the vocabulary adapted to the description of the domain,
defining an indexation model and identifying the suitable way to access
images depending on the degree of precision of the designer’s need.The
defined vocabulary, when associated with images, must be organised
according to five visual properties. Finally, we proposed a particular way to
access images without introducing text.All these principles have been
applied to the specific domain of architecture built in wood.
We carried out a final experiment to evaluate if the proposed method
satisfies a designer in a design situation.Thanks to the analyses of the
sketches produced by the designer during the final experiment, we were
able to evaluate in concrete terms the help that this kind of research can
bring to designers during their creation activity.
Future works are divided in two parts.The first one concerns the
application of these principles to other technical domains belonging to the
architecture field.The second one concerns the use of semantic networks
as another kind of navigation to access images.
References
1. Heylighen.A, IN CASE OF ARCHITECTURAL DESIGN, critique and praise of case-based design in architecture, Katholieke Universiteit Leuven Ed, may 2000.
2. Lebahar, J.-Ch.. Le dessin d’architecte, simulation graphique et réduction d’incertitude.Parenthèse Ed, Marseille, 1997.
3. Cabré,M.T, Terminology:Theory, Methods and Applications, Juan C. Sager ed,Amsterdam, Netherlands, 1999.
4. Dauzats, M, Le thésaurus de l’image: étude des langages documentaires pourl’audiovisuel, Paris, 1994.
5. Rosch, E.. Natural Categories, Cognitive Psychology, Vol. 4, issue 3, elsevier Ed1973.
6. Bignon, J.C., Halin, G., Nakapan,W, Building Product Information Search byImages. Proceedings of the 5th International Conference in Design and DecisionSupport Systems in Architecture, Nijkerk,The Netherlands, pp. 47-61, 2000.
7. G. Duffing,Thematico-Visual Image Retrieval: How to Deal With Partially IndexedCorpora, Internet Imaging II, San Jose, California, 2001.
8. Noy Fridman, N and Mc Guinness, D.L, Ontology Development 101:A Guide toCreating Your First Ontology, technical report, Stanford Knowledge systems
242 Sabrina Kacher, Gilles Halin, Jean-Claude Bignon and Pascal Humbert
laboratory and stanford medical Informatics, Stanford, USA, 2001.
9. Hudon, M.. Le thesaurus, conception, elaboration, gestion., Asted ed, Ottawa,Canada, 1994
10. Aitchison, J and Gilbrichrist,A, Thesaurus construction: a practical manual, SecondEdition, London, 1987.
11. Sabrina Kacher, Jean-Claude Bignon, Gilles Halin and Pascal Humbert,A semanticindexation method to assist with the architectural design process by images,Sigradi 2004,VIII Iberoamerican Congress of Digital Graphic, november 2004, PortoAlegre, Brasil.
12. Sabrina Kacher, Jean-Claude Bignon, Gilles Halin,A method to index images inthe wooden architecture domain,Terms hierarchy and weight given to terms,DDSS 2004, 7th International Conference on Design & Decision Support Systems inArchitecture and Urban Planning 2 – 5 July, 2004, Netherlands
13. Denis, M. On figurative components of mental representations, In F. Klix, J. Hoffmann,&E. van der Meer Ed, cognitive research in psychology,Amsterdam, 1982.
14. Reed, S. K. COGNITION théories et applications, De Boeck Ed, Paris, 1999.
15. Ching Francis D-K, Architecture: form, space and order, New-York:Van NostrandReinhold, New york, USA, 1996.
16. Joly martine, Introduction à l’anayse de l’image, Nathan Université Ed, 1993, Paris
17. Yadolah Dodge, Statistique Dictionnaire encyclopédique, Dunod Ed, 1993, Paris.
243A method for Constructing a Reference Image Database to Assist with Design Process.
Application to the Wooden Architecture Domain
Sabrina Kacher, Gilles Halin, Jean-Claude Bignon and Pascal Humbert.
CRAI (research centre in Architecture and Engineering),Architecture school
of Nancy, 2, rue Bastien Lepage – B.P. 435 54001 Nancy CEDEX, France.
Top Related